Nickel Oxide Reduction by Hydrogen: Kinetics and ... - ACS Publications

9 downloads 5352 Views 7MB Size Report
Jun 22, 2015 - combustion, which is a novel approach for green power generation.6−8 NiO reduction and the subsequent Ni behavior in hydrogen are of ...
Article pubs.acs.org/JPCC

Nickel Oxide Reduction by Hydrogen: Kinetics and Structural Transformations Khachatur V. Manukyan,*,†,‡ Arpi G. Avetisyan,‡ Christopher E. Shuck,§ Hakob A. Chatilyan,‡ Sergei Rouvimov,# Suren L. Kharatyan,‡,⊥ and Alexander S. Mukasyan*,§ †

Department of Physics, University of Notre Dame, Notre Dame, Indiana 46556, United States Laboratory of Kinetics of SHS Processes, Institute of Chemical Physics NAS of Armenia, Yerevan 0014, Armenia § Department of Chemical and Biomolecular Engineering, University of Notre Dame, Notre Dame, Indiana 46556, United States # Department of Electrical Engineering, University of Notre Dame, Notre Dame, Indiana 46556, United States ⊥ Department of Chemistry, Yerevan State University, Yerevan 0025, Armenia ‡

W Web-Enhanced Feature * S Supporting Information *

ABSTRACT: We studied the reduction kinetics of bulk NiO crystals by hydrogen and the corresponding structural transformations in the temperature range of 543−1593 K. A new experimental approach allows us to arrest and quench the reaction at different stages with millisecond time resolution. Two distinctive temperature intervals are found where the reaction kinetics and product microstructures are different. At relatively low temperatures, 543−773 K, the kinetic curves have a sigmoidal shape with long induction times (up to 2000 s) and result in incomplete conversion. Low-temperature reduction forms a complex polycrystalline Ni/NiO porous structure with characteristic pore size on the order of 100 nm. No induction period was observed for the high-temperature conditions (1173−1593 K), and full reduction of NiO to Ni is achieved within seconds. An extremely fine porous metal structure, with pore size under 10 nm, forms during hightemperature reduction by a novel crystal growth mechanism. This consists of the epitaxial-like transformation of micrometersized NiO single crystals into single-crystalline Ni without any crystallographic changes, including shape, size, or crystal orientation. The Avrami nucleation model accurately describes the reaction kinetics in both temperature regimes. However, the structural transformations during reduction in both nanolevel and atomic level are very complex, and the mechanism relies on both nucleation and the critical diffusion length for outward diffusion of water molecules.



order,2 Avrami,27−29 geometrical construction,26 and Szekely− Evans30,31 were proposed to describe this reaction. Several in situ techniques were used to investigate the mechanism of NiO reduction. Richardson et al. studied2 the reduction of porous NiO particles by hydrogen using a timeresolved X-ray diffraction technique in the temperature range of 450−570 K. A three-step reaction mechanism was proposed: (i) nickel clusters form during an induction period; (ii) next, the reaction rate accelerates as the clusters grow; (iii) finally, the process settles to a pseudo-first-order reaction with respect to nickel. Jeangros et al. (573−923 K) used environmental transmission electron microscopy (TEM) coupled with electron energy loss spectroscopy for in situ study of hydrogen reduction of NiO,11 On the basis of a model-based fitting procedure, they suggested that the Avrami (nucleation and nuclei growth) model describes the reaction. This assumes that hydrogen molecules predominantly dissociate on nickel atoms

INTRODUCTION Nickel is an important industrial catalyst and is typically produced by reduction of nickel oxide (NiO).1−5 This metal is also a promising oxygen carrier in chemical looping combustion, which is a novel approach for green power generation.6−8 NiO reduction and the subsequent Ni behavior in hydrogen are of practical importance in solid oxide fuel cells,9,10 where it determines the structure of the electronic conductor on the anode side. It is also a model process for the investigation of oxide reduction mechanisms2,11−15 due to its single-stage transformation, as compared to Fe2O316−19 and CuO.20−22 Despite extensive investigations of NiO reduction by hydrogen since the pioneering work23 of Benton and Emmett in 1924, many questions remain about its kinetics and the related structural transformation mechanism. For example, the reported activation energy varies by an order of magnitude, between 10 and 150 kJ/mol.2,11,24−26 This large variation may be explained by the lack of consistency between NiO samples, characterization techniques, and other experimental parameters. Furthermore, different kinetic models, including reaction © 2015 American Chemical Society

Received: May 5, 2015 Revised: June 21, 2015 Published: June 22, 2015 16131

DOI: 10.1021/acs.jpcc.5b04313 J. Phys. Chem. C 2015, 119, 16131−16138

Article

The Journal of Physical Chemistry C

Figure 1. Microstructure of NiO: surface of wire (A), bright field TEM image (B) of a thin slice of oxide layer, diffraction pattern (C), and a highresolution TEM image (D).

that are surrounded by oxygen vacancies.32 Rodrı ́guez et al. used synchrotron time-resolved X-ray adsorption and diffraction techniques to study the reduction of single-crystalline and polycrystalline NiO between 523 and 620 K.1 They found a strong correlation between the rate of reduction and the amount of oxygen defects in the metal oxide structure. It was suggested that the presence of oxygen vacancies increases the adsorption energy of H2 on the oxide surface, as well as decreases the energy barrier for hydrogen dissociation. H atoms then diffuse to the NiO reaction sites, resulting in a rupturing of the Ni−O bond with simultaneous desorption of a water molecule. Studies using optical microscopy by Utigard et al.33 led to the conclusion that the geometric construction, or shrinking core, model represents the reduction kinetics of NiO granules in the range of 100−500 μm. This model assumes that, after the reduction reaction has begun, the solid particles consist of a nonreacted core encased in a uniform layer of the reaction product. These two regions are sharply separated from each other by a well-defined interface. As the reaction proceeds, diffusion of hydrogen through the Ni layer becomes the ratelimiting step. Later, Plascencia and Utigard showed34 that the reduction of large porous NiO particles can be described by the grain model proposed by Szekely et al.35 In this empirical model, a particle is considered to be an agglomeration of individual grains, and each of them undergoes a microscopic shrinking core reduction. Depending on the process conditions and characteristics of NiO, either chemistry (low temperatures, high porosities) or internal diffusion (high temperatures, low porosities) controls the reaction kinetics. Recently, Hidayat et al. identified several distinctive porous microstructures,36 during the reduction of bulk dense NiO samples. Jeangros et al. also reported11 the macropore (50−100 nm) formation during early stages of NiO nanoparticle reduction by hydrogen. However, toward the end of reduction, the pores disappeared. The severe collapse of pores has also been reported during the reduction of hollow NiO nanoparticles at 550−870 K.37−39 Evidently, hydrogen reduction of nickel oxide is a complex process, which includes a variety of topological and morphological factors. It is also clear that the microstructural evolution during reduction and its relation to the kinetic models is not well understood. Herein, we report the results of reduction kinetics of NiO by hydrogen in a wide temperature range (543−1593 K). A novel experimental method permits arresting and rapidly quenching (with 104 K/s cooling rate) the reaction at different stages within milliseconds. This approach, coupled with ex situ microscopy techniques, allows us to measure the reaction rate and precisely track the ongoing

structural transformation. The mechanism of NiO reduction by hydrogen is suggested based on direct correlations between reaction kinetics and the observed structural transformations at the micro- and nanolevels and atomic levels.



EXPERIMENTAL SECTION The experiments were performed using partially oxidized nickel wires (Supporting Figure S1). To prepare these specimens an initial Ni wire (purity = 99.995%, 100 μm, and 8.5 cm in length) was heated by direct current to 1573 K and retained at this temperature for 150 s in air. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images (Figure 1A,B) illustrate that the formed NiO layer is polycrystalline with grain size in the range of 1−3 μm. Analysis of the selected area electron diffraction pattern (Figure 1C) and high-resolution TEM images (Figure 1D) reveal that the individual NiO grains are pore-free single crystals. A reaction chamber equipped with infrared-transparent windows, high-speed solar cell-based temperature sensors, a power source, and a PC-controlled unit (Supporting Figure 2) was used in this hydrogen reduction experiment (Supporting Figure 2). The reaction setup allows rapid (up to 4.5 × 105 K/ s) controllable heating of the wires (Supporting Figure 3) and continuous data (temperature, electrical power, resistivity of the wire) acquisition with a frequency of 10 kHz40,41 (see details in the Supporting Information). These NiO/Ni wires were inserted into the chamber, which was then evacuated to 10−3 kPa, purged with pure hydrogen, and finally filled with H2 to the desired pressure from 1.0 to 7.0 kPa. Following each experiment, no pressure change was observed. Utilizing the internal Ni metallic core, the wires were controllably heated by direct current to the desired temperature (543−1593 K). The termination of electrical current through the thin wire allows rapid arresting of the process with a quenching rate of ∼104 K/ s (Supporting Figure 3). This feature permits both arresting of the reaction and preservation of the sample structure with millisecond time resolution. A Mettler-Toledo balance (measuring accuracy of 10−5 g) was used to obtain the sample weight at different stages of the process to determine the degree of conversion (α) during reduction reaction as α = −Δm /Δmo

(1)

where Δmo is the mass change of Ni wire after initial oxidation and Δm is the subsequent mass change of the Ni/NiO wire during the reduction reaction. Obtained kinetic curves were analyzed using custom Matlab nonlinear least-squares regression analysis. A variety of models, totaling 18, were tested, including reaction order, geometrical 16132

DOI: 10.1021/acs.jpcc.5b04313 J. Phys. Chem. C 2015, 119, 16131−16138

Article

The Journal of Physical Chemistry C

Figure 2. Kinetics of NiO reduction at 543 K (A), 608 K (B), 1503 K (C), and 1593 K (D) at 6.67 kPa (solid points) and 1.33 kPa (open points) hydrogen pressures.

produce a clean cross section without milling artifacts. Thereafter, the electron beam was used to image the side wall of the trench at an angle of 52° parallel to the side wall. The TEM samples were also prepared using a Helios NanoLab 600 system by making cross-sectional slices from the top surfaces of the wires. The slice and view (S&V, FEI) software package was used on the Helios NanoLab 600 to collect a series of sequential images. A voltage of 30 kV with a milling current of 9.7 pA was used to make a series of 100 images with ∼10 μm width and ∼7.5 μm depth. These images were taken in series after milling 10 nm between each image, leading to a particle volume of ∼75 μm3 imaged throughout and prepared for pore characterization.

contraction (R1, R2, R3), one-, two-, or three-dimension diffusion-limited (D1, D2, D3), Avrami−Erofeev (AE) nucleation models, and others.26,42 The fitting parameters including pre-exponential factor (k0), activation energy (Ea), and exponential term (n) were varied for each model, with the two-parameter models having only k0 and Ea fit. All models were analyzed in series, with the varying parameters ranging over 20 log units of perturbation over the local attempt-space. The results with the highest goodness of fit (R2) were compared across models. Each kinetic data set was analyzed both individually and in every inclusive range. The inclusive data sets with the best goodness of fits were interpreted to be within the same kinetic regime. SEM and TEM were employed to characterize the composition and morphology of the reaction products, as well as the atomic structure of the materials. SEM analysis was conducted on a Magellan 400 (FEI, USA) with a resolution of 0.6 nm. The Magellan 400 is equipped with an energy dispersive X-ray spectrometer (EDS, Bruker) with energy resolution of 123 eV. A Titan 80-300 (FEI, USA) transmission electron microscope, with resolution of 0.136 nm in scanning TEM mode and about 0.1 nm information limit in highresolution TEM mode, was used. The Titan is equipped with an energy dispersive X-ray spectroscopy (EDS, Oxford Inca) system with spectral energy resolution of 130 eV. A Helios NanoLab 600 system (FEI) with dual electron/ion beam was used to produce cross sections of reacted wires by ion milling with a gallium ion beam. First, a ∼2 μm thick layer of platinum in a 15 μm × 1 μm rectangular area was deposited onto the selected sample. Next, a trench with ∼10 μm width, ∼15 μm length, ∼7.5 μm depth, and a 45° base angle was milled on the particle surface under an accelerating voltage of 10 kV with a milling current of 27 nA. Then, a 40 nm thick layer was milled into the side wall of the trench under an accelerating voltage of 10 kV and a milling current of 700 pA to



RESULTS

Below we report the kinetics for NiO reduction by hydrogen at different temperatures (e.g., 543, 608, 1503, and 1593 K) at constant H2 pressure (e.g., 1.33 and 6.67 kPa), as well as the characteristic structures of the reaction media on micro- and nanoscales and atomic scales at different stages of the process. Reaction Kinetics. Kinetics curves of NiO reduction by hydrogen, which represent the degree of conversion (α) versus time (t) at different temperatures and constant H2 pressure, are presented in Figure 2. It can be seen that, at 543 K (Figure 2A) and 608 K (Figure 2B), the kinetic curves have a sigmoidal shape with low reaction rates at initial stages, followed by a significant acceleration of the process after some critical time (induction period). Supporting Table 1 shows that the induction period depends on the temperature and pressure and can be as long as 2000 s. Another important observation is that full conversion cannot be achieved. The kinetics curves at 1503 K (Figure 2C) and 1593 K (Figure 2D) possess different behaviors. No induction period is detected and the degree of conversion monotonically increases with time in a nonlinear (parabolic) fashion. Figure 1D also 16133

DOI: 10.1021/acs.jpcc.5b04313 J. Phys. Chem. C 2015, 119, 16131−16138

Article

The Journal of Physical Chemistry C demonstrates that the hydrogen pressure has essentially no influence on the reaction rates. Finally, complete reduction of NiO could be achieved within 10−30 s. The results of the kinetic nonlinear least-squares analysis indicate that only the nth order Avrami-Erofeev (AEn) model is relevant for the low-temperature reduction (1173 K) kinetics. The AEn model has the form αNiO = 1 − exp( −ko exp( −Ea /(RT ))t n)

(2)

where the Avrami exponent (n) varies from 0.5 to 4, with each value describing a specific type of kinetic limitation; values below 1 indicate a diffusion-limited reaction, with integers 1 and above corresponding to growth in n dimensions. The D1 model takes the form αNiO = (k 0 exp( −Ea /(RT ))t )(1/2)

(3)

The results of model-based fitting analysis for 608 and 1503 K temperatures are shown in panels B and C of Figure 2, respectively. Kinetic data at 608 K fits with R2 = 0.98 by using the AEn model with a value of exponent n = 0.58 (Figure 2B). The kinetic data for 1503 K can be also accurately (R2 = 0.99) fit by the AEn model with n = 0.95 (Figure 2C), as well as by the D1 model (R2 = 0.95). All other models show less fitting accuracy. Figure 3 shows the n parameter as a function of temperature over the entire studied range. In low temperatures (543−773

Figure 4. Microstructure of surfaces (A, C) and cross sections (B, D) of reacted samples at early stages of reduction.

summarizes the microstructure analysis of the samples reduced at early stages of the reaction for both temperatures. A porous structure is predominantly formed on the crystal vertices close to the boundaries between the NiO grains after the sample had been heated at 608 K for 50 s (Figure 4A). However, this structure accounts for